combination of images may be used for forest classification. The short revisit time also strongly supports the change detection
applications. The objective of this paper is to show first results of delineated
forest areas from multi-temporal Sentinel-1 data from an Austrian study site. The results are validated with a reference
forest mask derived from airborne full-waveform laser scanning data.
2. STUDY AREA AND DATA
2.1 Study area
The northern part of the federal state Burgenland in Austria was selected as study area
Figure 1
. In addition to the flat area around the Lake Neusiedl, this region contains two hilly areas
with the highest elevation of 748 m and 484 m a.s.l., respectively. Both hilly regions are almost completely covered
with mixed forest deciduous and coniferous trees with about 400 km
2
each, while there are also several smaller forest areas up to 40 km
2
outside the closed forests. The rest of the study area is rather flat, open land or urban area, with elevations ranging
from 110 m to 260 m a.s.l., while in the eastern part the lake Neusiedl is located.
Figure 1: Overview of the study area. The ArcGIS Online basemap orthophoto is overlaid with the normalized DSM ,
derived from ALS data. The cells where the normalized DSM is smaller than 1.5 m are plotted as transparent.
2.2
Data:
Sentinel-1A IW Ground Range Detected GRD Level-1 product was used in this study. The IW data have spatial resolution of
20m with the pixel spacing of 10m. Over the study area, regular coverage in VV and VH polarization is available since October
2014 with a temporal resolution of 4 days on average. As the volume scattering at C-Band in forests representing
permanently vegetated areas and dense agricultural crops representing seasonally vegetated areas may cause similar
backscatter values over these two classes in summer time acquisitions and thus make the two classes less separable
Balzter et al., 2015, only the winter period 1
st
December 2014 to 31
st
M arch 2015 data were selected. All acquisitions from the specified time frame were used, regardless of the
environmental conditions snow, precipitation, temperature or differences in orbit and thus in local incidence angles. On
average, 30 acquisitions are available for each pixel within the study area.
In addition to Sentinel-1 data, full-waveform airborne laser scanning ALS data are available. The ALS data were collected
in April 2010 with Riegl LM S-Q560 and LM S- Q680 sensors under the leaf-off and snow-free conditions. The ALS data cover
the whole study area with an average point density in the non- overlapping areas of 4 pointsm
2
, while recording up to 15 echoes per single laser shoot. The laser footprint of the recorded
data was not larger than 60 cm in diameter. The ALS data are used for deriving a reference forest mask used for validating the
Sentinel-1 derived forest area. 2.3
Reference forest mask
For the generation of the reference forest mask the approach from Eysn et al. 2012 is applied. This comprehensive
approach is based on ALS data by considering the criterions tree height, crown coverage CC and the minimum area and width.
The criterion of land use is not considered in this approach. Based on the slope adaptive echo ratio map, which describes the
transparency for laser beams of the top most surface, it is possible to differentiate between buildings and trees. For the
calculation of the CC the approach from Eysn et al. 2011 is applied, which uses clear geometric definition of the CC and
works on a similar way than it is the case for manual assessment of the CC based on aerial orthophotos. This approach uses the
area of the convex hull of three neighbouring trees as reference unit and thus overcomes limitations from e.g. pure moving
window approaches such as smoothing effects along the forest border or the dependency on the kernel size and shape of the
moving window. The entire workflow is implemented in the software package Opals Pfeifer et al., 2014.
This contribution has been peer-reviewed. The double-blind peer-review was conducted on the basis of the full paper. doi:10.5194isprsannals-III-7-227-2016
228
Figure 2: Subset of the derived forest mask with a minimum mapping unit of 500 m², crown coverage ≥30, minimum height
3 m and minimum width of 10 m. In the background an ArcGIS Online basemap orthophoto is shown.
3. METHODS
